Studying Signal Transduction Networks with a Probabilistic Boolean Network Approach

In recent years, various modelling approaches in systems biology have been applied for the study and analysis of signal transduction networks. However, each modelling approach has its inherent advantages and disadvantages, so the choice has to be made based on research objectives and types of data. In this PhD dissertation, we propose probabilistic Boolean network (PBN) as one of the suitable modelling approaches for studying signal transduction networks with steady-state data. The steady-state distribution of molecular states in PBN can be correlated to the steady-state proteomic profiles generated from wet-lab experiments. In addition, the relevance of interactions within signalling networks can be assessed through the optimised selection probabilities. These features make PBNs ideal for describing the properties of signal transduction networks at steady-state with some uncertainty on network topologies.
To investigate the applicability of PBNs for the study of signal transduction networks, we developed optPBN, an optimisation and analysis toolbox in the PBN framework. We demonstrated that optPBN can be applied to optimise a large-scale apoptotic network with 96 nodes and 105 interactions. Also, it allows for network contextualisation in a physiological context of primary hepatocytes through the analysis on optimised selection probabilities. Similarly, we also applied optPBN to study deregulated signal transduction networks in pathological contexts, i.e. the PDGF signalling in gastrointestinal stromal tumour (GIST) and the L-plastin signalling in breast cancer cell lines. By integrating prior information on network topology from literature with context-specific experimental data, contextualised PBNs can be derived which in turn provide additional insights into biological systems such as the importance of certain crosstalk interactions and the comparative signal flows at steady-state in non-metastatic versus metastatic cancer cell lines.
In addition to the applications on fundamental research, we also explored the applications of PBNs in a pharmaceutical setting where detailed mechanistic models are usually used.
Here, we applied optPBN as a tool for network ontextualisation. A proof-of-concept example on a small model demonstrated that optPBN helped to pre-select the suitable network structure according to the provided experimental data prior to the building and optimisation of detailed mechanistic models. Such application is foreseen to be applied in a pharmaceutical setting and to explore additional applications such as combinatorial drugs’ effect and toxicity screening.